SCAMTRACE — Exposing the Connections Behind Every Scam

Inspiration

Online scams are becoming increasingly sophisticated. From fake banking messages and fraudulent investment opportunities to phishing links and impersonation attacks, people encounter suspicious digital communications every day.

One of the biggest problems is that scam reports are often isolated. Someone might report a suspicious phone number, another person might report a fraudulent UPI ID, and someone else might receive a similar phishing message. However, these incidents may be connected, and users have very few ways to discover those relationships.

We wanted to address this gap.

The idea behind SCAMTRACE was to move beyond simply detecting whether something looks suspicious. We wanted to build a platform that could investigate potential connections between scam reports, identify recurring patterns, and transform isolated information into actionable intelligence.

What it does

SCAMTRACE is an AI-powered scam intelligence and investigation platform that analyzes suspicious digital activity and visualizes potential relationships between reported incidents.

Its key capabilities include:

  • Multi-format Scam Analysis: Analyze suspicious messages, URLs, phone numbers, email addresses, UPI IDs, and screenshots.
  • AI-Powered Detection: Identify potential scam categories, suspicious indicators, and psychological manipulation techniques.
  • Explainable Risk Assessment: Generate risk scores alongside explanations, supporting evidence, and uncertainties.
  • Interactive Intelligence Graph: Visualize relationships between suspicious entities, incidents, domains, phone numbers, payment identifiers, and potential scam campaigns.
  • Incident Investigation: Explore related reports, examine evidence, maintain investigation notes, and track suspicious activity.
  • Threat Intelligence Dashboard: Monitor emerging scam patterns, incident trends, and suspicious entity activity.

The central idea is simple: instead of investigating every suspicious message independently, SCAMTRACE helps uncover the bigger picture.

How we built it

We designed SCAMTRACE as a full-stack application with a modular architecture.

  • Frontend: React, TypeScript, Vite, Tailwind CSS, and shadcn/ui for a responsive cybersecurity intelligence interface.
  • Visualization: Recharts for analytical dashboards and React Flow for interactive investigation graphs.
  • Backend: Python and FastAPI for API requests, investigation workflows, and data processing.
  • Database: PostgreSQL and SQLAlchemy for storing incidents, entities, relationships, evidence, and investigations.
  • AI and Intelligence: A combination of rule-based detection, AI-assisted scam classification, entity extraction, and semantic similarity matching to identify potentially related reports.

We also incorporated a demonstration environment with fictional scam scenarios to showcase the investigation workflow without depending entirely on external data providers.

Challenges we faced

One of our biggest challenges was designing a system that could identify relationships between seemingly unrelated scam reports.

A suspicious phone number, a payment identifier, and a phishing domain are different types of information. Connecting them meaningfully requires structured entity extraction, relationship modelling, and evidence-based reasoning.

Another challenge was balancing AI-generated insights with reliability. We wanted SCAMTRACE to explain why something appeared suspicious rather than blindly produce a risk score. We therefore designed the system to distinguish evidence, similarity, confidence, and uncertainty.

Building an interactive intelligence graph also presented challenges. The graph needed to support entity selection, relationship exploration, filtering, and path highlighting while remaining understandable as the network grew.

Finally, we had to balance an ambitious feature set with the limited development time of a hackathon. We prioritized the core investigation experience and designed the architecture to support further expansion.

What we learned

Developing SCAMTRACE helped us understand that effective AI applications require much more than integrating an LLM.

We learned the importance of combining deterministic rules with AI reasoning, structuring unorganized information into meaningful entities, modelling relationships between records, and presenting complex analytical results through intuitive visualizations.

We also gained practical experience with full-stack development, API integration, database modelling, graph visualization, and designing explainable AI-powered workflows.

Most importantly, we learned that detecting an individual threat is only part of the problem. Understanding how multiple threats might be connected can provide much greater investigative value.

What's next?

We envision expanding SCAMTRACE into a collaborative, continuously improving scam intelligence network.

Our future plans include:

  • Integrating trusted external threat intelligence feeds.
  • Improving semantic matching across large collections of scam reports.
  • Developing stronger campaign and relationship discovery.
  • Introducing privacy-preserving community reporting.
  • Adding real-time threat notifications.
  • Supporting additional languages and regional scam patterns.
  • Exploring integrations with cybersecurity organizations and financial fraud reporting ecosystems.

Our long-term vision is to make scam intelligence more accessible by helping individuals and investigators move from isolated suspicious messages to a broader understanding of potential fraud networks.


SCAMTRACE — Exposing the connections behind every scam.

Built With

Share this project:

Updates

Submission history